1 Comparing Humans and AI Agents Javier Insa-Cabrera 1, David L. Dowe 2, Sergio España-Cubillo 1, M.Victoria Hernández-Lloreda 3, José Hernández Orallo 1 1.Departament de Sistemes Informàtics i Computació, Universitat Politècnica de València, Spain. 2.Computer Science & Software Engineering, Clayton School of I.T., Monash University, Clayton, Victoria, 3800, Australia. 3.Departamento de Metodología de las Ciencias del Comportamiento, Universidad Complutense de Madrid, Spain AGI’ Mountain View, California, 3-7 August 2011

2 Outline Measuring intelligence universally Precedents Test setting and administration Agents and interfaces Results Discussion

Measuring intelligence universally 3 Project: anYnt (Anytime Universal Intelligence)  Any kind of system (biological, non-biological, human)  Any system now or in the future.  Any moment in its development (child, adult).  Any degree of intelligence.  Any speed.  Evaluation can be stopped at any time.  Can we construct a ‘universal’ intelligence test?

4  Imitation Game “Turing Test” (Turing 1950):  It is a test of humanity, and needs human intervention.  Not actually conceived to be a practical test for measuring intelligence up to and beyond human intelligence.  CAPTCHAs (von Ahn, Blum and Langford 2002):  Quick and practical, but strongly biased.  They evaluate specific tasks.  They are not conceived to evaluate intelligence, but to tell humans and machines apart at the current state of AI technology.  It is widely recognised that CAPTCHAs will not work in the future (they soon become obsolete). Precedents

5  Tests based on Kolmogorov Complexity (compression-extended Turing Tests, Dowe 1997a-b, 1998) (C-test, Hernandez-Orallo 1998).  Look like IQ tests, but formal and well-grounded.  Exercises (series) are not arbitrarily chosen.  They are drawn and constructed from a universal distribution, by setting several ‘levels’ for k: Precedents  However...  Some relatively simple algorithms perform well in IQ-like tests (Sanghi and Dowe 2003).  They are static (no planning abilities are required).

6  Universal Intelligence (Legg and Hutter 2007): an interactive extension to C-tests from sequences to environments. = performance over a universal distribution of environments.  Universal intelligence provides a definition which adds interaction and the notion of “planning” to the formula (so intelligence = learning + planning).  This makes this apparently different from an IQ (static) test. Precedents π μ riri oioi aiai

7  A definition of intelligence does not ensure an intelligence test.  Anytime Intelligence Test (Hernandez-Orallo and Dowe 2010):  An interactive setting following (Legg and Hutter 2007) which addresses:  Issues about the difficulty of environments.  The definition of discriminative environments.  Finite samples and (practical) finite interactions.  Time (speed) of agents and environments.  Reward aggregation, convergence issues.  Anytime and adaptive application.  An environment class  (Hernandez-Orallo 2010) (AGI-2010). Precedents In this work we perform an implementation of the test and we evaluate humans and a reinforcement learning algorithm with it, as a proof of concept.

Test setting and administration 8  Implementation of the environment class :  Spaces are defined as fully connected graphs.  Actions are the arrows in the graphs.  Observations are the ‘contents’ of each edge/cell in the graph.  Agents can perform actions inside the space.  Rewards:  Two special agents Good ( ⊕ ) and Evil ( ⊖ ), which are responsible for the rewards. Symmetric behaviour, to ensure balancedness.

9  We randomly generated only 7 environments for the test:  Different topologies and sizes for the patterns of the agents Good and Evil (which provide rewards).  Different lengths for each session (exercise) accordingly to the number of cells and the size of the patterns.  The goal was to allow for a feasible administration for humans in about minutes. Test setting and administration

10  An AI agent: Q-learning  A simple choice. A well-known algorithm.  A biological agent: humans  20 humans were used in the experiment  A specific interface was developed for them, while the rest of the setting was equal for both types of agents.  Agents and interfaces

11  Experiments were paired.  Results show that performance is fairly similar. Results

12  Analysis of the effect of complexity :  Complexity is approximated by using LZ (Lempel-Ziv) coding to the string which defines the environment. Results  Lower variance for exercises with higher complexity.  Slight inverse correlation with complexity ( difficulty , reward  ).

13  Not many studies comparing human performance and machine performance on non-specific tasks.  The environment class here has not been designed to be anthropomorphic.  The AI agent (Q-learning) has not been designed to address this problem.  The results are consistent with the C-test (Hernandez-Orallo 1998) and with the results in (Sanghi & Dowe 2003), where a simple algorithm is competitive in regular IQ tests. Discussion

14  The results show this is not a universal intelligence test.  The use of an interactive test has not changed the picture from the results in the C-test.  What may be wrong?  A problem of the current implementation. Many simplifications made.  A problem of the environment class. Both this and the C-test used an inappropriate reference machine.  A problem of the environment distribution.  A problem with the interfaces, making the problem very difficult for humans.  A problem of the theory.  Intelligence cannot be measured universally.  Intelligence is factorial. Test must account for more factors.  Using algorithmic information theory to precisely define and evaluate intelligence may be insufficient. Discussion

Thank you! Some pointers: Project: anYnt (Anytime Universal Intelligence) Have fun with the test